1 /* $OpenBSD: pax.c,v 1.28 2005/08/04 10:02:44 mpf Exp $ */
2 /* $NetBSD: pax.c,v 1.5 1996/03/26 23:54:20 mrg Exp $ */
5 * Copyright (c) 1992 Keith Muller.
6 * Copyright (c) 1992, 1993
7 * The Regents of the University of California. All rights reserved.
9 * This code is derived from software contributed to Berkeley by
10 * Keith Muller of the University of California, San Diego.
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 * 3. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38 static const char copyright[] =
39 "@(#) Copyright (c) 1992, 1993\n\
40 The Regents of the University of California. All rights reserved.\n";
45 static const char sccsid[] = "@(#)pax.c 8.2 (Berkeley) 4/18/94";
47 static const char rcsid[] = "$OpenBSD: pax.c,v 1.28 2005/08/04 10:02:44 mpf Exp $";
52 #include <sys/types.h>
53 #include <sys/param.h>
56 #include <sys/resource.h>
67 static int gen_init(void);
70 * PAX main routines, general globals and some simple start up routines
74 * Variables that can be accessed by any routine within pax
76 int act = DEFOP; /* read/write/append/copy */
77 FSUB *frmt = NULL; /* archive format type */
78 int cflag; /* match all EXCEPT pattern/file */
79 int cwdfd; /* starting cwd */
80 int dflag; /* directory member match only */
81 int iflag; /* interactive file/archive rename */
82 int kflag; /* do not overwrite existing files */
83 int lflag; /* use hard links when possible */
84 int nflag; /* select first archive member match */
85 int tflag; /* restore access time after read */
86 int uflag; /* ignore older modification time files */
87 int vflag; /* produce verbose output */
88 int Dflag; /* same as uflag except inode change time */
89 int Hflag; /* follow command line symlinks (write only) */
90 int Lflag; /* follow symlinks when writing */
91 int Xflag; /* archive files with same device id only */
92 int Yflag; /* same as Dflag except after name mode */
93 int Zflag; /* same as uflag except after name mode */
94 int zeroflag; /* use \0 as pathname terminator */
95 int vfpart; /* is partial verbose output in progress */
96 int patime = 1; /* preserve file access time */
97 int pmtime = 1; /* preserve file modification times */
98 int nodirs; /* do not create directories as needed */
99 int pmode; /* preserve file mode bits */
100 int pids; /* preserve file uid/gid */
101 int rmleadslash = 0; /* remove leading '/' from pathnames */
102 int exit_val; /* exit value */
103 int docrc; /* check/create file crc */
104 char *dirptr; /* destination dir in a copy */
105 char *ltmfrmt; /* -v locale time format (if any) */
106 char *argv0; /* root of argv[0] */
107 sigset_t s_mask; /* signal mask for cleanup critical sect */
108 FILE *listf; /* file pointer to print file list to */
109 char *tempfile; /* tempfile to use for mkstemp(3) */
110 char *tempbase; /* basename of tempfile to use for mkstemp(3) */
113 * PAX - Portable Archive Interchange
115 * A utility to read, write, and write lists of the members of archive
116 * files and copy directory hierarchies. A variety of archive formats
117 * are supported (some are described in POSIX 1003.1 10.1):
119 * ustar - 10.1.1 extended tar interchange format
120 * cpio - 10.1.2 extended cpio interchange format
121 * tar - old BSD 4.3 tar format
122 * binary cpio - old cpio with binary header format
123 * sysVR4 cpio - with and without CRC
125 * This version is a superset of IEEE Std 1003.2b-d3
127 * Summary of Extensions to the IEEE Standard:
129 * 1 READ ENHANCEMENTS
130 * 1.1 Operations which read archives will continue to operate even when
131 * processing archives which may be damaged, truncated, or fail to meet
132 * format specs in several different ways. Damaged sections of archives
133 * are detected and avoided if possible. Attempts will be made to resync
134 * archive read operations even with badly damaged media.
135 * 1.2 Blocksize requirements are not strictly enforced on archive read.
136 * Tapes which have variable sized records can be read without errors.
137 * 1.3 The user can specify via the non-standard option flag -E if error
138 * resync operation should stop on a media error, try a specified number
139 * of times to correct, or try to correct forever.
140 * 1.4 Sparse files (lseek holes) stored on the archive (but stored with blocks
141 * of all zeros will be restored with holes appropriate for the target
143 * 1.5 The user is notified whenever something is found during archive
144 * read operations which violates spec (but the read will continue).
145 * 1.6 Multiple archive volumes can be read and may span over different
147 * 1.7 Rigidly restores all file attributes exactly as they are stored on the
149 * 1.8 Modification change time ranges can be specified via multiple -T
150 * options. These allow a user to select files whose modification time
151 * lies within a specific time range.
152 * 1.9 Files can be selected based on owner (user name or uid) via one or more
154 * 1.10 Files can be selected based on group (group name or gid) via one o
156 * 1.11 File modification time can be checked against existing file after
157 * name modification (-Z)
159 * 2 WRITE ENHANCEMENTS
160 * 2.1 Write operation will stop instead of allowing a user to create a flawed
161 * flawed archive (due to any problem).
162 * 2.2 Archives written by pax are forced to strictly conform to both the
163 * archive and pax the specific format specifications.
164 * 2.3 Blocking size and format is rigidly enforced on writes.
165 * 2.4 Formats which may exhibit header overflow problems (they have fields
166 * too small for large file systems, such as inode number storage), use
167 * routines designed to repair this problem. These techniques still
168 * conform to both pax and format specifications, but no longer truncate
169 * these fields. This removes any restrictions on using these archive
170 * formats on large file systems.
171 * 2.5 Multiple archive volumes can be written and may span over different
173 * 2.6 A archive volume record limit allows the user to specify the number
174 * of bytes stored on an archive volume. When reached the user is
175 * prompted for the next archive volume. This is specified with the
176 * non-standard -B flag. The limit is rounded up to the next blocksize.
177 * 2.7 All archive padding during write use zero filled sections. This makes
178 * it much easier to pull data out of flawed archive during read
180 * 2.8 Access time reset with the -t applies to all file nodes (including
182 * 2.9 Symbolic links can be followed with -L (optional in the spec).
183 * 2.10 Modification or inode change time ranges can be specified via
184 * multiple -T options. These allow a user to select files whose
185 * modification or inode change time lies within a specific time range.
186 * 2.11 Files can be selected based on owner (user name or uid) via one or more
188 * 2.12 Files can be selected based on group (group name or gid) via one o
190 * 2.13 Symlinks which appear on the command line can be followed (without
191 * following other symlinks; -H flag)
193 * 3 COPY ENHANCEMENTS
194 * 3.1 Sparse files (lseek holes) can be copied without expanding the holes
195 * into zero filled blocks. The file copy is created with holes which are
196 * appropriate for the target filesystem
197 * 3.2 Access time as well as modification time on copied file trees can be
198 * preserved with the appropriate -p options.
199 * 3.3 Access time reset with the -t applies to all file nodes (including
201 * 3.4 Symbolic links can be followed with -L (optional in the spec).
202 * 3.5 Modification or inode change time ranges can be specified via
203 * multiple -T options. These allow a user to select files whose
204 * modification or inode change time lies within a specific time range.
205 * 3.6 Files can be selected based on owner (user name or uid) via one or more
207 * 3.7 Files can be selected based on group (group name or gid) via one o
209 * 3.8 Symlinks which appear on the command line can be followed (without
210 * following other symlinks; -H flag)
211 * 3.9 File inode change time can be checked against existing file before
212 * name modification (-D)
213 * 3.10 File inode change time can be checked against existing file after
214 * name modification (-Y)
215 * 3.11 File modification time can be checked against existing file after
216 * name modification (-Z)
218 * 4 GENERAL ENHANCEMENTS
219 * 4.1 Internal structure is designed to isolate format dependent and
220 * independent functions. Formats are selected via a format driver table.
221 * This encourages the addition of new archive formats by only having to
222 * write those routines which id, read and write the archive header.
227 * parse options, set up and operate as specified by the user.
228 * any operational flaw will set exit_val to non-zero
229 * Return: 0 if ok, 1 otherwise
233 main(int argc, char **argv)
239 * On some systems, stderr is not a constant, so we initialize listf
240 * immediately to emulate the behavior.
245 * Keep a reference to cwd, so we can always come back home.
247 cwdfd = open(".", O_RDONLY);
249 syswarn(1, errno, "Can't open current working directory.");
254 * Where should we put temporary files?
256 if ((tmpdir = getenv("TMPDIR")) == NULL || *tmpdir == '\0')
258 tdlen = strlen(tmpdir);
259 while (tdlen > 0 && tmpdir[tdlen - 1] == '/')
261 tempfile = malloc(tdlen + 1 + sizeof(_TFILE_BASE));
262 if (tempfile == NULL) {
263 paxwarn(1, "Cannot allocate memory for temp file name.");
267 memcpy(tempfile, tmpdir, tdlen);
268 tempbase = tempfile + tdlen;
272 * parse options, determine operational mode, general init
275 if ((gen_init() < 0) || (tty_init() < 0))
279 * select a primary operation mode
289 if (gzip_program != NULL)
290 errx(1, "can not gzip while appending");
306 * when interrupted we try to do whatever delayed processing we can.
307 * This is not critical, but we really ought to limit our damage when we
308 * are aborted by the user.
314 sig_cleanup(int which_sig)
316 /* XXX signal races */
319 * restore modes and times for any dirs we may have created
320 * or any dirs we may have read. Set vflag and vfpart so the user
321 * will clearly see the message on a line by itself.
324 if (which_sig == SIGXCPU)
325 paxwarn(0, "Cpu time limit reached, cleaning up.");
327 paxwarn(0, "Signal caught, cleaning up.");
338 * general setup routines. Not all are required, but they really help
339 * when dealing with a medium to large sized archives.
345 struct rlimit reslimit;
346 struct sigaction n_hand;
347 struct sigaction o_hand;
350 * Really needed to handle large archives. We can run out of memory for
351 * internal tables really fast when we have a whole lot of files...
353 if (getrlimit(RLIMIT_DATA , &reslimit) == 0){
354 reslimit.rlim_cur = reslimit.rlim_max;
355 (void)setrlimit(RLIMIT_DATA , &reslimit);
359 * should file size limits be waived? if the os limits us, this is
360 * needed if we want to write a large archive
362 if (getrlimit(RLIMIT_FSIZE , &reslimit) == 0){
363 reslimit.rlim_cur = reslimit.rlim_max;
364 (void)setrlimit(RLIMIT_FSIZE , &reslimit);
368 * increase the size the stack can grow to
370 if (getrlimit(RLIMIT_STACK , &reslimit) == 0){
371 reslimit.rlim_cur = reslimit.rlim_max;
372 (void)setrlimit(RLIMIT_STACK , &reslimit);
376 * not really needed, but doesn't hurt
378 if (getrlimit(RLIMIT_RSS , &reslimit) == 0){
379 reslimit.rlim_cur = reslimit.rlim_max;
380 (void)setrlimit(RLIMIT_RSS , &reslimit);
384 * Handle posix locale
386 * set user defines time printing format for -v option
388 ltmfrmt = getenv("LC_TIME");
391 * signal handling to reset stored directory times and modes. Since
392 * we deal with broken pipes via failed writes we ignore it. We also
393 * deal with any file size limit through failed writes. Cpu time
394 * limits are caught and a cleanup is forced.
396 if ((sigemptyset(&s_mask) < 0) || (sigaddset(&s_mask, SIGTERM) < 0) ||
397 (sigaddset(&s_mask,SIGINT) < 0)||(sigaddset(&s_mask,SIGHUP) < 0) ||
398 (sigaddset(&s_mask,SIGPIPE) < 0)||(sigaddset(&s_mask,SIGQUIT)<0) ||
399 (sigaddset(&s_mask,SIGXCPU) < 0)||(sigaddset(&s_mask,SIGXFSZ)<0)) {
400 paxwarn(1, "Unable to set up signal mask");
403 memset(&n_hand, 0, sizeof n_hand);
404 n_hand.sa_mask = s_mask;
406 n_hand.sa_handler = sig_cleanup;
408 if ((sigaction(SIGHUP, &n_hand, &o_hand) < 0) &&
409 (o_hand.sa_handler == SIG_IGN) &&
410 (sigaction(SIGHUP, &o_hand, &o_hand) < 0))
413 if ((sigaction(SIGTERM, &n_hand, &o_hand) < 0) &&
414 (o_hand.sa_handler == SIG_IGN) &&
415 (sigaction(SIGTERM, &o_hand, &o_hand) < 0))
418 if ((sigaction(SIGINT, &n_hand, &o_hand) < 0) &&
419 (o_hand.sa_handler == SIG_IGN) &&
420 (sigaction(SIGINT, &o_hand, &o_hand) < 0))
423 if ((sigaction(SIGQUIT, &n_hand, &o_hand) < 0) &&
424 (o_hand.sa_handler == SIG_IGN) &&
425 (sigaction(SIGQUIT, &o_hand, &o_hand) < 0))
428 if ((sigaction(SIGXCPU, &n_hand, &o_hand) < 0) &&
429 (o_hand.sa_handler == SIG_IGN) &&
430 (sigaction(SIGXCPU, &o_hand, &o_hand) < 0))
433 n_hand.sa_handler = SIG_IGN;
434 if ((sigaction(SIGPIPE, &n_hand, &o_hand) < 0) ||
435 (sigaction(SIGXFSZ, &n_hand, &o_hand) < 0))
440 syswarn(1, errno, "Unable to set up signal handler");